
Claude Interactive Connectors: Your Apps Now Run Inside Claude
Agentic AIAnthropic now lets partners run their full apps inside Claude, so a Confluence page or Notion database opens beside the chat and Claude edits it live. For enterprise IT and the developers building MCP servers, that moves Claude from an assistant that calls tools to the surface where the work happens, and it raises the stakes on who controls what Claude can change.
What Happened
The Decoder reports that Anthropic added Claude Sheets, Claude Whiteboards and Claude Data, formerly Dashboards, alongside upgraded interactive connectors. Anthropic’s announcement describes Sheets as collaborative spreadsheets with formulas across multiple tabs, Whiteboards as a shared space Claude can summarize after a meeting, and Data as a tool that connects to Snowflake and Salesforce to build dashboards. Claude Meetings, which will take notes on calls, is coming “soon.”
The connector change is the bigger one. Anthropic’s connector post says that, where a connector supports it, a tool can open beside the conversation or full screen, and users approve Claude’s edits “the same way you approve a connector’s actions.” It is rolling out on the web and desktop app on all plans, starting with Asana, Canva, Carta, Dropbox, monday.com, Smartsheet, Vercel and Webflow. Figma, Harvey and Shopify are building integrations.
Why It Matters
Interactive connectors run on MCP Apps, the open Model Context Protocol extension that lets a server bring its own interface into an AI application. Claude now supports more of it, including app tools and embedded sites, in public preview for developers. Anthropic stresses that an MCP App “isn’t limited to Claude” and can run in any client that supports the extension. That matters for builders: the interface work is portable, not a Claude-only plugin.
Distribution is already large. Anthropic says Claude Marketplace lists more than 9,000 connectors and plugins, and plugins from Salesforce (beta), NetSuite and Workday (beta) ship with built-in skills for sellers, accountants and HR teams. Anthropic’s own examples are concrete: its legal team built a Slack-based review tool that cut marketing reviews from three days to 24 hours, and a Gusto lawyer built a contract tool that took turnaround “from over a week to hours.”
What This Means for Builders
Embedding the full app means Claude acts with the user’s own signed-in session, so the governance controls deserve as much attention as the features. On Enterprise plans, admins choose who can use which connectors, set spend limits, and can restrict a connector’s tools, so Claude “might be able to read Slack but not send messages.” Each cloud session runs in its own isolated virtual machine with connector credentials kept outside it, and audit logs feed a Compliance API that works with more than 100 tools, including Splunk and CrowdStrike. Auto-connect, in beta, connects employees to approved apps on first use.
This continues the push Stack Archive covered in Claude Cowork’s take on mundane enterprise work: Anthropic is aiming at non-engineers who never open a terminal.
The So What
Enterprise admins should review connector tool permissions before interactive connectors reach their users, starting with write actions in tools like Confluence, Notion and Box. Teams that already run an MCP server for an internal app should prototype an MCP App interface now, since the same build can serve Claude and any other client that adopts the extension.
Source: The Decoder
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →